Performance Assessment for Automatic Generation Control via Dynamic Models Identified From Extracted Data Segments

Abstract

Automatic generation control (AGC) systems in thermal generation units keep the generated active power tracking the AGC commands dispatched from dispatching departments of power grids. The AGC performance of generation units is crucial for power grids to maintain their electrical energy balance and is of high concern to power plants and power grids. The problem is to estimate the ramp rate and static deviation as two AGC performance metrics from desired and generated active powers. This paper proposes an AGC performance assessment method to address two challenges in estimating the two performance metrics. One challenge is that not all data segments of the desired active power with amplitude variations are suitable for performance assessment. Another challenge is that severe noise induces uncertainties in the estimates of performance metrics. For the first challenge, the proposed method extracts step-pattern data segments, from which dynamic models are identified and performance metrics are estimated from model step responses. For the second challenge, uncertainties of the estimated performance metrics are quantified by confidence intervals obtained from the dynamic models with surrogate parameters. The benefits of the proposed method over the existing ones include: (1) invalid estimates are avoided by selecting step-pattern data segments for AGC performance assessment; (2) the root mean squared estimation errors are reduced by more than 60% in typical examples; (3) the uncertainties in the estimates are quantified by their confidence intervals. Numerical and industrial examples are provided to illustrate the effectiveness and benefits of the proposed method.